Predicting a Customer’s Lifetime Value
A Chicago retailer wanted to use their customer data to improve targeting, increase ROI, and predict customer lifetime value (CLV), but their data was decentralized and unreliable.
2nd Watch’s retail and marketing analytics team built a centralized data platform to quickly access a reliable single view of the customer. With the data foundation in place, we built an automated machine learning solution to predict CLV based on previous purchase behavior and other demographic information.
The result was a modern data solution that allowed the marketing team to identify predictors, better tailor marketing activities, and make data-driven marketing decisions.
- Which customers are the most valuable? Which is the best marketing channel to target high value customers? These are the questions many marketing professionals have but few have the ability to quickly and accurately answer.
- When a Chicago retailer, rich in customer data, wanted to improve their customer targeting, boost ROI, and more accurately predict customer lifetime value (CLV), they needed a modern data solution to make it happen.
- With this goal in mind, they turned to the retail and marketing analytics team at 2nd Watch for a machine learning solution to predict customer lifetime value.
The Business Challenges
Like many organizations, this retailer struggled with decentralized and unreliable data. Customer data was held in a wide variety of online platforms, e-commerce systems, and an in-store POS. A centralized data platform was needed before any data driven insights could be found.
Once the foundational data platform was put into place, the next step was an advanced analytics and machine learning solution that enabled our client to instantly analyze data and use new insights to help them predict the marketing efforts that would bring in the most valuable customers for the lowest cost, thus boosting their ROI.
The 2nd Watch Solution
Part 1: The Centralized Data Platform
The first step to trustworthy data was to create a centralized data platform. To do this, the 2nd Watch team ingested data from an in-store POS, e-commerce system, Google Analytics, and Facebook Insights using a FiveTran data pipeline into a Snowflake cloud- based data warehouse. Once ingested, the data was cleaned up to make it usable for analytics and machine learning.
While each step may sound simple, the 2nd Watch team’s knowledge and approach to building long-term data solutions means that the overall platform will withstand growth, alert users when something is wrong, and accommodate changing needs.
Part 2: A Retail-Focused Machine Learning Solution
With the data foundation in place, an automated machine learning platform, DataRobot, was used to run machine learning models and produce a predicted CLV based on a customer’s previous purchase behavior and other demographic information. Additionally interactive Tableau dashboards were created to answer questions such as:
• Which customers have the highest CLV?
• What is the best channel to target high value customers?
• What does our overall marketing performance look like by channel?
• What activities are most likely to result in a conversion?
• What is the average cost per conversion?
By implementing a consolidated data warehouse and machine learning, 2nd Watch was able to identify predictors so the client could better tailor marketing activities. Not only was their holistic view of marketing vastly improved, but they were able to drill down to a single customer to anticipate future spending habits and increase conversion through better targeting.